Story Tree - Autonomous Hierarchical Backlog Manager

Purpose

Maintain a self-managing tree of user stories where:

• Each node represents a story at some level of granularity
• Each node has a capacity (target number of child nodes)
• The skill autonomously identifies under-capacity nodes and generates stories to fill them
• Git commits are analyzed to mark stories as implemented
• Higher-level nodes are prioritized when under capacity

Design rationale: If instructions seem counter-intuitive, consult references/rationales.md before changing approach.

When NOT to Use

• Creating 1-3 specific stories manually
• Non-hierarchical backlogs
• Projects without git history
• Real-time task tracking
• Detailed implementation planning (use superpowers:writing-plans)
• Non-software projects

Storage

Database: .claude/data/story-tree.db Schema: references/schema.sql

Ensure .gitignore includes: *.db with exception !.claude/data/story-tree.db

Environment Requirements

CRITICAL: The sqlite3 CLI command is NOT available in most environments. Always use Python's sqlite3 module:

python -c "
import sqlite3
conn = sqlite3.connect('.claude/data/story-tree.db')
cursor = conn.cursor()
cursor.execute('YOUR SQL HERE')
print(cursor.fetchall())
conn.close()
"

Script locations: All scripts are in .claude/skills/story-tree/scripts/ (NOT project root scripts/):

• Tree visualization: .claude/skills/story-tree/scripts/tree-view.py

Do NOT:

• Use sqlite3 CLI command directly
• Assume scripts/ means project root
• Try multiple approaches hoping one works

Autonomous Operation

When user says "update story tree" or "generate stories":

1. Run complete workflow (Steps 1-7) without asking permission
2. Generate stories based on git analysis and priority algorithm
3. Output complete report when finished
4. Ask for clarification ONLY when: over-capacity detected, multiple equal priorities, or ambiguous git history

Auto-Update on Staleness

On ANY invocation, check lastUpdated in metadata. If >3 days old, run full update first.

Workflow Steps

Step 1: Initialize/Load Database

If .claude/data/story-tree.db doesn't exist:

mkdir -p .claude/data

Then use Python (NOT sqlite3 CLI):

python -c "
import sqlite3
import json
from pathlib import Path

# Get project name
try:
    project_name = json.load(open('package.json')).get('name', 'Software Project')
except:
    project_name = 'Software Project'

# Read and execute schema
schema_path = '.claude/skills/story-tree/references/schema.sql'
schema = Path(schema_path).read_text()

conn = sqlite3.connect('.claude/data/story-tree.db')
conn.executescript(schema)

# Initialize root node
conn.execute('''
    INSERT INTO story_nodes (id, title, description, status, created_at, updated_at)
    VALUES ('root', ?, 'Project root', 'active', datetime('now'), datetime('now'))
''', (project_name,))

conn.execute('''INSERT INTO story_paths (ancestor_id, descendant_id, depth) VALUES ('root', 'root', 0)''')
conn.execute('''INSERT INTO metadata (key, value) VALUES ('version', '2.4.0')''')
conn.execute('''INSERT INTO metadata (key, value) VALUES ('lastUpdated', datetime('now'))''')
conn.commit()
conn.close()
print('Database initialized')
"

Step 2: Analyze Git Commits

Use incremental analysis from checkpoint:

python -c "
import sqlite3
import subprocess

conn = sqlite3.connect('.claude/data/story-tree.db')
cursor = conn.cursor()
cursor.execute(\"SELECT value FROM metadata WHERE key = 'lastAnalyzedCommit'\")
row = cursor.fetchone()
last_commit = row[0] if row else None
conn.close()

# Check if commit exists
if last_commit:
    result = subprocess.run(['git', 'cat-file', '-t', last_commit], capture_output=True)
    if result.returncode != 0:
        last_commit = None

if last_commit:
    cmd = ['git', 'log', f'{last_commit}..HEAD', '--pretty=format:%h|%ai|%s', '--no-merges']
else:
    cmd = ['git', 'log', '--since=30 days ago', '--pretty=format:%h|%ai|%s', '--no-merges']

result = subprocess.run(cmd, capture_output=True, text=True)
print(result.stdout)
"

Match commits to stories using keyword similarity (see references/sql-queries.md#pattern-matching).

Step 3: Identify Priority Target

Excluded statuses: concept, rejected, wishlist, epic, deprecated, infeasible, bugged

Priority algorithm - find under-capacity nodes, shallower first:

SELECT s.*,
    (SELECT COUNT(*) FROM story_paths WHERE ancestor_id = s.id AND depth = 1) as child_count,
    (SELECT MIN(depth) FROM story_paths WHERE descendant_id = s.id) as node_depth,
    COALESCE(s.capacity, 3 + (SELECT COUNT(*) FROM story_paths sp
         JOIN story_nodes child ON sp.descendant_id = child.id
         WHERE sp.ancestor_id = s.id AND sp.depth = 1
         AND child.status IN ('implemented', 'ready'))) as effective_capacity
FROM story_nodes s
WHERE s.status NOT IN ('concept', 'rejected', 'wishlist', 'epic', 'deprecated', 'infeasible', 'bugged')
  AND (SELECT COUNT(*) FROM story_paths WHERE ancestor_id = s.id AND depth = 1) <
      COALESCE(s.capacity, 3 + (SELECT COUNT(*) FROM story_paths sp
           JOIN story_nodes child ON sp.descendant_id = child.id
           WHERE sp.ancestor_id = s.id AND sp.depth = 1
           AND child.status IN ('implemented', 'ready')))
ORDER BY node_depth ASC
LIMIT 1;

Dynamic capacity: effective_capacity = capacity_override OR (3 + implemented/ready children)

Step 4: Generate Stories (Max 3 per node)

Story format:

### [ID]: [Title]

**As a** [specific user role]
**I want** [specific capability]
**So that** [specific benefit]

**Acceptance Criteria:**
- [ ] [Specific, testable criterion]
- [ ] [Specific, testable criterion]

**Related context**: [Git commits or patterns]

New nodes start with status: 'concept'. When user explicitly requests "generate stories for [node-id]", create with status: 'approved' instead.

Step 5: Insert Stories

-- Insert story (capacity NULL = dynamic)
INSERT INTO story_nodes (id, title, description, status, created_at, updated_at)
VALUES (:new_id, :title, :description, 'concept', datetime('now'), datetime('now'));

-- Populate closure table
INSERT INTO story_paths (ancestor_id, descendant_id, depth)
SELECT ancestor_id, :new_id, depth + 1
FROM story_paths WHERE descendant_id = :parent_id
UNION ALL SELECT :new_id, :new_id, 0;

Step 6: Update Metadata

INSERT OR REPLACE INTO metadata (key, value) VALUES ('lastUpdated', datetime('now'));
INSERT OR REPLACE INTO metadata (key, value) VALUES ('lastAnalyzedCommit', :newest_commit);

Step 7: Output Report

# Story Tree Update Report
Generated: [ISO timestamp]

## Current Tree Status
- Total nodes: [N]
- Implemented: [N] ([%]%)
- In progress: [N] ([%]%)

## Git Commits Analyzed
[List matched commits]

## Priority Target
**Node**: [ID] "[Title]" - Level [N], [children]/[capacity]

## Generated Stories
[Full story format for each]

## Tree Visualization
[Run tree-view.py and present output]

Tree Visualization

Script path: .claude/skills/story-tree/scripts/tree-view.py (NOT scripts/tree-view.py)

python .claude/skills/story-tree/scripts/tree-view.py --show-capacity

The script automatically handles UTF-8 encoding on Windows. Use --force-ascii only if Unicode rendering fails.

Status symbols:

User Commands

Quality Checks

Before outputting stories, verify:

• Each story has clear basis in commits or gap analysis
• Stories are specific and actionable
• Acceptance criteria are testable
• No duplicates
• User story format complete

Common Mistakes (STOP Before Making These)

Red Flags - If you're doing any of these, STOP:

• Running a command without checking if it exists first
• Assuming any path is relative to project root
• Using shell commands when Python is specified
• Iterating through failed approaches

References

• references/schema.sql - Database schema (source of truth)
• references/sql-queries.md - All SQL query patterns
• references/common-mistakes.md - Error prevention
• references/rationales.md - Design decisions
• references/epic-decomposition.md - Epic/wishlist workflow